Tangent Works was one of three worldwide final match winners at GEFCom2017, the Global Energy Forecasting Competition. At the GEFCom2017 competition, 177 teams were registered, with only 73 of those teams producing results. The TIM technology proved that it can compete with—and surpass— some of the best energy forecasting academic teams in the world.
TANGENT WORKS is a European machine learning company specializing in advanced forecasting for the energy sector. The company builds on a strong academic foundation, with a focus on scalable industrial solutions and services. Its main product, TIM (Tangent Information Modeler), is a fully automatic model-building engine that makes state of the art academic modelling available to all companies.
Since the inception of our company by a team of mathematicians, statisticians, machine learning experts, engineers and business people, we have had a simple yet grand vision.
To create a product which would combine our deep and extensive knowledge in the field of differential and information geometry into a unique product which could be automatically deployed to address forecast needs in the energy sector. Since our creation, we‘ve teamed up with Microsoft and Julia Computing to make our technology accessible and therefore remove the burden of modelling by mathematicians.
Our vision was to create a product which would behave almost like an excellent data science team. A team would receive a dataset and would create a model which then can be used to generate a forecast on demand. Our vision has one small difference - we expect machines to come out with better or at least the same quality results in seconds instead of days.
Last year we successfully made our product available via API which allows access to our cloud-based Tangent Information Modeller. At the same time, we successfully commenced several commercial deployments of our product with our technology partners and customers.
We were confident of the fact that we have built a world-class automatic forecasting engine and this has been confirmed by our customers‘ numerous industry benchmarks. However, as we are a relatively small and new company that is active in the energy sector, which is still a traditional industry, we have faced the challenge of our performance claims being accepted. This is due to the fact that our work with our customers occurs under strict NDA's with confidential datasets.
Due to this, we began searching for an independent benchmark institution which compares commercial forecasting tools. Last year, we came across GEFCOM 2014, and decided to test our engine‘s ex-post in this competition. The results were excellent, we achieved top place in all tracks of GEFCOM 2014. From these successful results, we decided to enter GEFCOM 2017, where we achieved finally 1st place in the classification and 2nd place in the final round.
GEFCOM 2017 was mainly about electricity and did not cover Intraday, Day Ahead/Next day and several days ahead scenario forecasting, which is where TANGENT really excels, despite this we wanted to benchmark our commercially available engine with top class academic research teams.
The purpose of this simple google site pages is to provide some resources about how our automated model building engine was used to achieve top places with Gefcom 2014 & Gefcom 2017 datasets.